基于內(nèi)容的新聞鏡頭分類技術(shù)研究

1、天津大學碩士學位論文基于內(nèi)容的新聞鏡頭分類技術(shù)研究姓名：楊益錚申請學位級別：碩士專業(yè)：信號與信息處理指導教師：蘇育挺20080501ABSTRACTWiththedevelopmentofintemetandcomputertechnologymoreandmoremultimediainformation，especiallydigitalvideocometopeople’SliveItisanimportanttaskforUSt

2、outilizetheinformationeffectivelyTofacilitatepeoplegettingtheinformationtheywant，buildingasystemforvideoindexing，browsingandsearchhasbeentheworkofresearchersAsabasicunitofvideo，shotcannotonlybedividedintoframesbutalsocom

3、posestory，whichisveryimportantforvideoanalysisVideoshotclassificationtechnologybridgesthesemanticgapbetweenlowlevelfeaturesandhighlevelconcepts，andprovidessupportforthevideoindexandretrievalNewsvideohasaclearstructurewhi

4、chmakesitplayanimportantroleinvideocontentanalysisInthisdissertation，newsshotsarecategorizedtosixsemantictypes：Commercial，StillImage，Anchorperson，ReporterMonologueandOthersThe“Others”aretherestshotsexceptthefirstfivetype

5、sinthenewsvideoThetypesofCommercial，StillImage，AnchorpersonandOthersareidentifiedrespectivelywithfeaturesofthemselvesAndthereporterandmonologueshotsaredistinguishedbyconditionalrandomfields(CRFs)model，inwhichthedetection

6、istransformedintosequencelabelingproblemTheexperimentalresultsdemonstratetheeffectivenessandhighperformanceofthemethodKEYWORDS：NewsVideo，ShotBoundaryDetection，ShotClassification，AdvertisementDetection，ConditionalRandomFi